The file must include a type field (LP, QP, or SDP)
along with the required matrices and parameters.
Required fields by type:
- LP:
type, A_d, b_d, c. Optional: G, h, rho, tau, confidence, p, x_ref.
- QP: Same as LP, plus
Q.
- SDP:
type, F_d (dict of expression matrices keyed "0","1",...), c. Optional: E (dict of numeric matrices), Q, rho, tau, confidence.
JSON format:
{
"type": "LP",
"A_d": [["-delta[0]", "0"], ["0", "-delta[0]"]],
"b_d": [["delta[0] - 1"], ["delta[0] - 1"]],
"G": [["-1", "0"], ["0", "-1"]],
"h": [["0"], ["0"]],
"c": [["1"], ["1"]],
"rho": "0",
"tau": "0",
"confidence": "0.01"
}
MAT format:
Each MATLAB variable maps to a field name. Variable names must match
the required fields above. Numeric matrices (e.g. G, h,
c, Q, E) are saved as standard arrays.
For fields with delta[i] expressions
(A_d, b_d, F_d),
use MATLAB cell arrays of strings.
% MATLAB example for LP:
type = 'LP';
c = [1; 1];
G = [-1 0; 0 -1]; h = [0; 0];
rho = 0; tau = 0; confidence = 0.01;
% delta expressions as cell arrays of strings:
A_d = {'-delta[0]', '0'; '0', '-delta[0]'};
b_d = {'delta[0] - 1'; 'delta[0] - 1'};
save('benchmark.mat', 'type', 'A_d', 'b_d', ...
'G', 'h', 'c', 'rho', 'tau', 'confidence')